Hi! Welcome to the 22nd edition of the TomorrowToday newsletter.

We’re here to decode the AI chaos so you don't have to. Think of us as your friendly neighbourhood tech translators - we cut through the chaos, translate the jargon, and spotlight new AI tools that matter for founders, builders, and curious minds.

Buckle up, because the future's moving fast and we're here to make sure you don't get left behind! ⚡

If you enjoyed today’s newsletter, please forward it to a friend & subscribe by following this link.

_{~5 mins read}

/xAI /Benchmark /Performance

No press releases. No hype. Nothing. Just results.

The xAI team quietly patched Grok-4-Fast... and the before/after is wild.

What changed? They improved how often Grok actually completes a task instead of giving up halfway through or returning an error. In "reasoning" mode (where the AI thinks step-by-step through complex problems), completion rates jumped from 77.5% to 94.1%. In standard "non-reasoning" mode, completions shot up to a staggering 97.9%.

What do these benchmarks measure? Completion rate is one of the most practical metrics in AI - it's the percentage of times the model successfully finishes what you asked it to do versus times it fails, times out, or returns an incomplete response. Think of it like asking your intern to draft 100 emails: would they successfully complete 77 of them, or 97?

How did they do it? All thanks to better "injected system prompts" - essentially, the hidden instructions that tell the AI how to approach problems before it even sees your question. xAI refined these background instructions to help Grok stay on track and finish tasks more reliably.

Why this matters: When you're building with AI, completion rates are everything. A chatbot that fails 22% of the time isn't production-ready. One that fails only 2% of the time? That's enterprise-grade. Grok just leapfrogged most of the competition without making a sound about it - no drama, just real improvements that make the model significantly more reliable in real-world applications.

/Google /No-Code

Google Labs just dropped Opal in 160+ countries, including South Africa - and it's a game-changer for anyone who's ever thought "I wish I could build that" but didn't know where to start.

Opal is a no-code platform that lets you describe what you want in plain English, and it builds a working AI mini-app for you. Chain together prompts, call AI models, add tools - all without touching a line of code. Just tell it what you need, and Opal translates your instructions into a visual workflow you can tweak and share immediately.

The platform runs on Google's Vertex AI and Gemini models, with built-in templates to get you started. Want to build a social media post generator? A meeting summariser? A custom productivity tool? You can have a working prototype in minutes.

For entrepreneurs and small teams, this removes one of the biggest barriers to AI adoption: technical expertise. If you can explain what you need, Opal can build it.

/Apple /Gemini /Siri

Apple is spending roughly $1 billion annually to license Google's Gemini for its long-delayed Siri overhaul, according to Bloomberg. But you won't hear them shouting about it.

Gemini will handle summarisation and multi-step planning within Siri, running on Apple's Private Cloud Compute infrastructure to keep user data private. The 1.2 trillion parameter model dwarfs the 150 billion parameters in Apple's current Intelligence model.

Apple tested models from OpenAI and Anthropic too, but landed on Gemini. Bloomberg reports the partnership is "unlikely to be promoted publicly" - Apple wants Google as a "behind-the-scenes" tech supplier while it builds its own capable model internally.

The new Siri could arrive as soon as next Spring. Apple views this as temporary, but considering the company's AI struggles and recent employee exodus, building their own solution from scratch feels like a long shot.

In this section, we aim to spotlight an incredible AI tool or use case and guide you on how you can try it.

🏗️ This week's challenge: Build Your Own AI App (Without Code)

Remember when building software required years of coding experience? Those days are over. With Google's newly launched Opal, you can go from idea to working app in under 10 minutes.

I decided to test it by building a callisthenics assistant - an AI tool that generates personalised bodyweight workout plans with visual demonstrations and form correction tips. From initial prompt to shareable app? Less than 10 minutes.

Here's how to set it up:

Step 1: Access Opal (it is free)

Step 2: Describe What You Want

Step 3: Test and Refine

Step 4: Share Your Creation

Pro Tips:

The real magic? Opal removes the friction between "wouldn't it be cool if" and "here's a working prototype." For South African founders and teams, this is your shortcut to validating ideas before investing in full development.

You spoke, we listened. “AI in Enterprise” is here to stay. In this section, we're spotlighting real businesses using AI to solve actual problems.

📊 This week: A&O Shearman Saves 2-3 Hours Weekly Per Lawyer

A&O Shearman, one of the world's elite law firms, deployed Harvey AI enterprise-wide across 4,000 staff in 43 jurisdictions. The result? They became Europe's "Most Innovative Law Firm" in 2024.

The Problem

Big Law has a dirty secret: highly-trained attorneys spend massive chunks of their billable hours on repetitive document work. Contract reviews, legal research, and precedent hunting - tasks that require legal knowledge but not strategic thinking. For a firm operating across dozens of countries, this inefficiency compounds quickly.

The Solution

A&O Shearman partnered with Harvey to build custom AI tools tailored to their specific workflows. Unlike generic chatbots, Harvey was trained on legal-specific data and integrated directly into their document management systems. Their flagship tool, ContractMatrix, is now used daily by 2,000 lawyers across the firm.

The Results

The numbers tell the story:

The Lesson

AI adoption in professional services isn't about replacing expertise - it's about eliminating the tedious work that buries it. A&O Shearman didn't implement AI to cut headcount; they did it so their lawyers could spend more time on the strategic work that justifies their fees.

For South African professional services firms - whether legal, accounting, or consulting - the playbook is clear: identify the high-volume, low-judgment tasks that consume your team's time, and let AI handle them. Your clients get better service, your team gets better work, and your margins improve.

AI is full of jargon, and we’re here to decode it. Each week, we’ll give you a plain-English definition of a buzzy term you’ve probably seen (but never fully understood).

Thanks for reading TomorrowToday! We’d love to hear from you:

➡️ What would you like us to cover next?
➡️ Have a tool or topic we should feature?

We’re building this with (and for) you. 🚀
See you next Tuesday 👋